Reconfigurable holographic surfaces (RHSs) constitute a promising technique of supporting energy-efficient communications. In this paper, we formulate the energy efficiency maximization problem of the switch-controlled RHS-aided beamforming architecture by alternately optimizing the holographic beamformer at the RHS, the digital beamformer, the total transmit power and the power sharing ratio of each user. Specifically, to deal with this challenging non-convex optimization problem, we decouple it into three sub-problems. Firstly, the coefficients of RHS elements responsible for the holographic beamformer are optimized to maximize the sum of the eigen-channel gains of all users by our proposed low-complexity eigen-decomposition (ED) method. Then, the digital beamformer is designed by the singular value decomposition (SVD) method to support multi-user information transfer. Finally, the total transmit power and the power sharing ratio are alternately optimized, while considering the effect of transceiver hardware impairments (HWI). We theoretically derive the spectral efficiency and energy efficiency performance upper bound for the RHS-based beamforming architectures in the presence of HWIs. Our simulation results show that the switch-controlled RHS-aided beamforming architecture achieves higher energy efficiency than the conventional fully digital beamformer and the hybrid beamformer based on phase shift arrays (PSA). Moreover, considering the effect of HWI in the beamforming design can bring about further energy efficiency enhancements.